Because my new way of dealing with any sort of negative emotions is to go full-on confrontational about it. Maybe one day I will find a balance, but for now this is positively euphoric.
Lol. That approach sounds like it has an interesting story or two behind it...and I can't say I wasn't somewhat amused by your openly blunt response, admittedly. Still, I hope that perhaps you'll be slightly more forgiving when you next set your sights upon me. I think I'm one of the more decent sorts of characters bouncing around the internet, all in all.
Nothing really concrete or easy to follow. The closest thing I have on hand would probably be the Octave scripts from my machine learning course, which are heavily based on the linear algebra taught in class and feature only as much documentation as I deemed helpful for the TA to follow along when marking.
I have a surprisingly high tolerance for not-easy-to-follow. ;3 Linear algebra, though, that's neat; I keep seeing references to its importance in the machine-learning field. I've not got that far yet; mostly I've been learning about the basics of calculus and integration in relation to physics studies I've been pursuing lately. No doubt I'm a lightweight at the number-crunching compared to you at this point. Do you find it to be a difficult subject? I understand computers do most of the heavy numerical lifting, but of course you have to formulate everything right or they spit out garbage, as always... X3
My math background isn't that strong, but I enjoy learning about the numerical side of nature nonetheless when I understand the physical ideas behind the numbers. I'm still figuring out what aspects of robotics and machine learning most attract me, but I've always loved robots of almost all sorts, and I know I find things like the Blue Brain project, and Kwabena Boahen's Neurogrid work, very fascinating. Do you have any special influences in that regard?
I'm especially keen to understand more about how neurons operate computationally, and the ways in which they can be reliably modeled in that respect, though I'm a ways off from any meaningful competence in that regard. Brains are just the most interesting piles of goop...it's hard not to get ahead of myself in my excitement, though, when thinking about the ways in which biological and technological solutions for observing and interacting with the world increasingly overlap.
Anyway, I'm curious now; where do you see yourself headed in the future, with regards to your current studies? Are there any memorable experiences you've had that particularly cemented those interests for you, as compared to other pursuits? (I hope I'm not being too presumptuous with my questions~)
I hope I don't sound like a hyperactive three year old when I butt into a conversation like this, but a friend and I have been an evolution sim that uses neural networks to simulate the 'brains' of each simulated 'bot'. It's very primitive, has been abandoned and can be vastly improved (like doing all the neural calculations on the GPU instead of it being bottlenecked by the CPU) and has been programmed in C++. (which, if you know basic control structures, is easy to follow-ish) If you'd be interested, I could probably upload the source for you to fool around in.
Oh and @MachineMuse, are animations allowed? (If so, would I make them as a single file that is about multipleof16*16 pixels... or would I have to make separate
files?)
Great mod! Really fun features. Some folks and I have been assembling themed suits:
Speed Tank - no defense, all offense. Super fast with flight mods.
Walking Mech - melee only, step assist, water electrolyzer, can't jump or swim (read: refuses to jump or swim).
Mobile Battery - Heavy armour, kinetic generator, carries a battery station (Advanced Power Management) to discharge at key sites. (High priority asset in convoy, defended by other players.)
It's way nerdy, but too much fun -- trying to tinker out of exploit-level quirks and keep things balanced. I've been reading PlayerTickHandler and thinking about some possible mass-related fun. e.g. higher weight penalties with added resistance to knockback -- sink faster/harder to swim -- less effective glider wings, but able to pull up (eventually) and glide without the walking penalty. It's just a bunch of force calculations, but still a pita. Also things like fixing free strafing in flight, but I need to do more reading.
Hehe, that's delightful! One of my major goals was to have different options for different playstyles, so I love hearing of different ways people specialize. Hopefully as MPS gets more refined and closer to something that could be considered a 1.0 release, the options for diversity will increase.
Lol. That approach sounds like it has an interesting story or two behind it...and I can't say I wasn't somewhat amused by your openly blunt response, admittedly. Still, I hope that perhaps you'll be slightly more forgiving when you next set your sights upon me. I think I'm one of the more decent sorts of characters bouncing around the internet, all in all.
I have a surprisingly high tolerance for not-easy-to-follow. ;3 Linear algebra, though, that's neat; I keep seeing references to its importance in the machine-learning field. I've not got that far yet; mostly I've been learning about the basics of calculus and integration in relation to physics studies I've been pursuing lately. No doubt I'm a lightweight at the number-crunching compared to you at this point. Do you find it to be a difficult subject? I understand computers do most of the heavy numerical lifting, but of course you have to formulate everything right or they spit out garbage, as always... X3
My math background isn't that strong, but I enjoy learning about the numerical side of nature nonetheless when I understand the physical ideas behind the numbers. I'm still figuring out what aspects of robotics and machine learning most attract me, but I've always loved robots of almost all sorts, and I know I find things like the Blue Brain project, and Kwabena Boahen's Neurogrid work, very fascinating. Do you have any special influences in that regard?
I'm especially keen to understand more about how neurons operate computationally, and the ways in which they can be reliably modeled in that respect, though I'm a ways off from any meaningful competence in that regard. Brains are just the most interesting piles of goop...it's hard not to get ahead of myself in my excitement, though, when thinking about the ways in which biological and technological solutions for observing and interacting with the world increasingly overlap.
Yeah, the math is kinda the hard part for me, too - it is actually very simple stuff in practice but wrapping my head around what it all means is proving a bit of a chore. But it's important, because translating everything into low-level operations means it can be optimized at that level, often in a way that makes the difference between time-complexity classes.
But when it comes to simulations? Computers and brains are so fundamentally different from each other that in my experience it's more effective to focus on things that each does well, and on interfacing them in ways that effectively bridge the gaps in their capabilities.
The simplest non-trivial brain-like data structure is the feed-forward neural network, so that's something to look into if you're just spreading your wings in this subject. The trick is, each 'layer' of neurons is actually just a matrix (representing a linear system) with a non-linear transformation function on the output, which is what makes it so you can start to do some really fancy math with it. And from there, you can start to develop it in ways that take advantage of the things a computer is good at, as opposed to a human.
Anyway, I'm curious now; where do you see yourself headed in the future, with regards to your current studies? Are there any memorable experiences you've had that particularly cemented those interests for you, as compared to other pursuits? (I hope I'm not being too presumptuous with my questions~)
I'm really more inspired by scifi on the topic and by the neuropsych courses I've taken which have shown me that the common conception 'We don't really understand how the brain works' is pretty far from the truth. I don't have any particularly strong focuses; the whole field is just so fascinating to me that every little development is exciting. I would more point to specific people than experiences, though, which I am not thrilled about doing on a public forum. I will at least mention my late grandmother, who sparked my interest in design and technology very early with her legos and engineer toys (Jacob's ladder, Newton's cradle, ring-and-string puzzles, etc.)
I hope I don't sound like a hyperactive three year old when I butt into a conversation like this, but a friend and I have been an evolution sim that uses neural networks to simulate the 'brains' of each simulated 'bot'. It's very primitive, has been abandoned and can be vastly improved (like doing all the neural calculations on the GPU instead of it being bottlenecked by the CPU) and has been programmed in C++. (which, if you know basic control structures, is easy to follow-ish) If you'd be interested, I could probably upload the source for you to fool around in.
Oh and @MachineMuse, are animations allowed? (If so, would I make them as a single file that is about multipleof16*16 pixels... or would I have to make separate
files?)
Yeah, animations in Minecraft are just tiled vertically now. Check out the lava texture in minecraft.jar if you need a demo.
I value forthrightness in discussion (though it can be disgruntling at times), and sometimes a measure of open controversy is a refreshing alternative to stifled politenesses. The world has enough legalese in it as it is. /:P
Yeah, the math is kinda the hard part for me, too - it is actually very simple stuff in practice but wrapping my head around what it all means is proving a bit of a chore. But it's important, because translating everything into low-level operations means it can be optimized at that level, often in a way that makes the difference between time-complexity classes.
Math is always very simple, in a way (or so I tell myself at the times I'm most stumped by it ). I mean, the expressions of a given function, etc, can be very complex and nuanced (not that I'm all that much of a mathematician), but in the end it all comes down to relatively simple logical foundations, stacked on top of each other in varyingly sophisticated - but unambiguous - ways, so far as I understand. I suppose it's more that human active memory chunking limitations are to blame. I know I run into them often enough. X3 I read some of Spivak's text on calculus and analysis, and while a lot of it is beyond my grasp, it definitely reinforced the inherrent logical rigour of the process in a rather adept way that I still try and hold on to in my other studies.
Personally, I usually find it's mostly a matter of grasping the deeper context, when it comes to having a sense of the numbers and operators. I think I have a rather visual bent to my mind; trigonometry is one of the easier things for me to grasp intuitively because it's easier to visualize the geometry in ways that clearly illustrate the core relationships. Finding those clarifying images is for me one of the more gratifying aspects of the learning process.
I don't know much about optimization concerns, seeing as I've never developed anything to the point where it matters, so I lack that insight. I'm aware of how important it can be in practice, however; one of the more memorable examples I know of is Quake's Fast InvSqrt() function in the days of yore. Could you elaborate on what you mean regarding "time-complexity classes"? Also, as for low-level operations, how low do most programmers typically go these days? Is there much place for assembly languages anymore, or would that go beyond what's typically expected in most development scenarios?
But when it comes to simulations? Computers and brains are so fundamentally different from each other that in my experience it's more effective to focus on things that each does well, and on interfacing them in ways that effectively bridge the gaps in their capabilities.
The simplest non-trivial brain-like data structure is the feed-forward neural network, so that's something to look into if you're just spreading your wings in this subject. The trick is, each 'layer' of neurons is actually just a matrix (representing a linear system) with a non-linear transformation function on the output, which is what makes it so you can start to do some really fancy math with it. And from there, you can start to develop it in ways that take advantage of the things a computer is good at, as opposed to a human.
Yes, I'm reasonably aware of how inherrently different brains and classical CPUs are, and I agree, a hybrid approach seems the best method all around. When I talked about convergence before, I was referring more to our growing capacity to bridge these two very distinctive approaches in a meaningful way. The Blue Brain project is sort of insane when you consider the hardware they're running it on; they're trying to emulate a massively parallel system with an inherrently sequential approach, and that's the main reason it's as expensive as it is. It's an inefficient realization, though of course it makes sense in context (economy, available tech, familiarity, etc). I'm very excited to see how memristors start to change things as they're developed and introduced commercially over the next few decades, since they'll seem to encourage some more innovative, inherrently parallel archetectures.
Regarding feed-forward neural networks; that makes sense, thanks for the insight. I've been reading a bit on how individual neurons can be modeled mathematically (the "leaky-integrate-and-fire" model, for example), but have not considered them in that broader, interlinked context yet. Out of curiousity, am I right to presume that in a simple feed-forward network as you describe, each layer of N neurons in a hierarchy of layers can be treated as a 1-by-N (horizontal) matrice (that is, can a single neuron in a layer be destribed by a single function, or is more than one function (columns) needed)? Or would that depend on the neural model used? (Either way, more to learn on my part~)
I'm really more inspired by scifi on the topic and by the neuropsych courses I've taken which have shown me that the common conception 'We don't really understand how the brain works' is pretty far from the truth. I don't have any particularly strong focuses; the whole field is just so fascinating to me that every little development is exciting. I would more point to specific people than experiences, though, which I am not thrilled about doing on a public forum. I will at least mention my late grandmother, who sparked my interest in design and technology very early with her legos and engineer toys (Jacob's ladder, Newton's cradle, ring-and-string puzzles, etc.)
Yes, it's hard to be selective in excitement, isn't it, with everything going on~ Especially in how each advance in a related field enhances the achievements of the others. And it's pretty incredible how far our understandings of brains have come. I was reading not long ago about a monkey that has been trained to control a third robotic arm through neural activity alone (back in 2010!); and of course there are exciting things happening in prosthetics more generally as well. And then of course there's all the buzz over drones, both autonomous and not. For better or worse, we live in interesting times.
Also, I get your concerns about posting that sort of stuff publicly, no problem. I simply enjoy talking about these topics, so please pardon me if I'm a bit overbearing. Regardless, I think it's great that your grandmother was a supportive influence as she was; that sort of rapport is important as a kid, when you're figuring out where to go in life. (I've still got all my own old lego buried somewhere...) I know one of my personal inspirations is Nikola Tesla. In addition to being a truly masterful engineer in general, he even dabbled in robotics and remote operation, well in advance of its widespread acceptance in more modern times. I sometimes wonder just what he might have been able to do with modern computer components, considering what he was able to do with the tools and materials of his own era...
Scifi is a strong influence for me too. I think a lot of the most interesting scifi stories are the ones that explore the psychology of thinking machines in various forms; comparing and contrasting them to other sorts of beings, and assessing how near or far they may fall to human and animal ways of thinking. Perhaps it's a somewhat gratuitous choice, but one of my favourite robots is the simply-titled Robo from Chrono Trigger, as he comes to certain conclusions about his own role and importance in the game(/story)'s driving conflict, as compared to those of his(?) human companions: Roles above those of a mere tool, servant, or obliged guardian. His wasn't the most active role (and as a story, Chrono Trigger has its weaknesses), but it brought forth some interesting questions upon reflection, for me anyway.
And of course there's the scifi greats like Asimov, complex characterizations like Star Trek's Data, and more epochryphal examples in older fiction, like golems and automatons; fantasy-robots of earlier times. Metropolis (both the original German film, and the Japanese anime inspired by it and Tezuka's work) also hang in my mind, for various reasons. Do you have any personal favourites of that sort that spring to mind?
Anyway, I think that whatever forms of robotic minds ultimately emerge from our endeavours as a species, humans will learn a tremendous amount about themselves in the process. I'm excited to see that happen, when it does.
(Oh lord, I've gone and made a gigantic text wall again. One day I'll figure this whole "brevity" thing out, and my replies to things will never exceed 250 words, and the internet will ring with bright rejoicing, and maybe I'll even get a firm, brisk handshake from the Singularity... =c= )
...If you'd be interested, I could probably upload the source for you to fool around in.
Lol, no worries. You know, that sounds quite neat, but it's probably more than I could do anything with at this point. I've only just begun looking at neural simulation in practice, and my programming skills are very modest as of yet. I've slowly learned that diving into things headfirst, without a solid grounding to stand on beforehand, generally results in a lot of mostly-useless floundering, and not much else. X3 I really appreciate the offer, though, and perhaps I'll ask you about it again sometime, when I'm more confident in my skills. Perhaps you should post it somewhere anyway, though, if you're keen to share? May be better than letting it moulder on a lonely hard drive. /
I value forthrightness in discussion (though it can be disgruntling at times), and sometimes a measure of open controversy is a refreshing alternative to stifled politenesses. The world has enough legalese in it as it is. /:P
Math is always very simple, in a way (or so I tell myself at the times I'm most stumped by it ). I mean, the expressions of a given function, etc, can be very complex and nuanced (not that I'm all that much of a mathematician), but in the end it all comes down to relatively simple logical foundations, stacked on top of each other in varyingly sophisticated - but unambiguous - ways, so far as I understand. I suppose it's more that human active memory chunking limitations are to blame. I know I run into them often enough. X3 I read some of Spivak's text on calculus and analysis, and while a lot of it is beyond my grasp, it definitely reinforced the inherrent logical rigour of the process in a rather adept way that I still try and hold on to in my other studies.
Personally, I usually find it's mostly a matter of grasping the deeper context, when it comes to having a sense of the numbers and operators. I think I have a rather visual bent to my mind; trigonometry is one of the easier things for me to grasp intuitively because it's easier to visualize the geometry in ways that clearly illustrate the core relationships. Finding those clarifying images is for me one of the more gratifying aspects of the learning process.
Hm, visual analogies sometimes help but I usually find sequential/verbal explanations and metaphors more useful. A visual analogy is meaningless without an understanding of what each part is trying to do. Although for SVMs it really does help...
I don't know much about optimization concerns, seeing as I've never developed anything to the point where it matters, so I lack that insight. I'm aware of how important it can be in practice, however; one of the more memorable examples I know of is Quake's Fast InvSqrt() function in the days of yore. Could you elaborate on what you mean regarding "time-complexity classes"? Also, as for low-level operations, how low do most programmers typically go these days? Is there much place for assembly languages anymore, or would that go beyond what's typically expected in most development scenarios?
For most programmers, it hardly matters at all, because you aren't dealing with time-critical code. Even in graphics programming, a loss of 2-3 fps on an average computer is hardly noticeable. But there is still a niche for it, because in some industries the amount of information you can process is directly correlated to the amount of money you can make, and even when not, there is still a significant cost savings if, for example, someone can make a few code changes at some bottleneck and avoid needing another server rack.
Time-complexity classes are a different issue, and one that's more related to classical computer science. It's really a pretty broad topic so I'll direct you to wikipedia, since this article is pretty well-written imo: http://en.wikipedia....Time_complexity
Yes, I'm reasonably aware of how inherrently different brains and classical CPUs are, and I agree, a hybrid approach seems the best method all around. When I talked about convergence before, I was referring more to our growing capacity to bridge these two very distinctive approaches in a meaningful way. The Blue Brain project is sort of insane when you consider the hardware they're running it on; they're trying to emulate a massively parallel system with an inherrently sequential approach, and that's the main reason it's as expensive as it is. It's an inefficient realization, though of course it makes sense in context (economy, available tech, familiarity, etc). I'm very excited to see how memristors start to change things as they're developed and introduced commercially over the next few decades, since they'll seem to encourage some more innovative, inherrently parallel archetectures.
Regarding feed-forward neural networks; that makes sense, thanks for the insight. I've been reading a bit on how individual neurons can be modeled mathematically (the "leaky-integrate-and-fire" model, for example), but have not considered them in that broader, interlinked context yet. Out of curiousity, am I right to presume that in a simple feed-forward network as you describe, each layer of N neurons in a hierarchy of layers can be treated as a 1-by-N (horizontal) matrice (that is, can a single neuron in a layer be destribed by a single function, or is more than one function (columns) needed)? Or would that depend on the neural model used? (Either way, more to learn on my part~)
I see. Yeah, there's certainly the parallel aspect, but there's more - for a really accurate brain simulation, you need to account for varying resource availability and environmental conditions in different regions of the brain, as well as spatial proximity and even the role of support cells (glia). There's probably more still.
As for the neural network: remember that each neuron (in the simplest model) is connected to each neuron in the following layer. So the actual state of the neuron can be a 1-by-N vector (a vector is just a matrix with a single row or column) representing the input or stimulation that each of them is receiving, but the actual data encoded in the neurons is M-by-N (where M is the number of neurons in the following layer). In other words: you can think of the rows as the 'weights' that a given 2nd-level neuron takes from each of the 1st-level neurons' outputs. So each output neuron has a value equal to a different weighted sum of the input neurons' values. Then you apply some transformation to it, probably a sigmoid:
To get a value between 0 and 1, representing the probability that that neuron is going to fire.
Yes, it's hard to be selective in excitement, isn't it, with everything going on~ Especially in how each advance in a related field enhances the achievements of the others. And it's pretty incredible how far our understandings of brains have come. I was reading not long ago about a monkey that has been trained to control a third robotic arm through neural activity alone (back in 2010!); and of course there are exciting things happening in prosthetics more generally as well. And then of course there's all the buzz over drones, both autonomous and not. For better or worse, we live in interesting times.
Also, I get your concerns about posting that sort of stuff publicly, no problem. I simply enjoy talking about these topics, so please pardon me if I'm a bit overbearing. Regardless, I think it's great that your grandmother was a supportive influence as she was; that sort of rapport is important as a kid, when you're figuring out where to go in life. (I've still got all my own old lego buried somewhere...) I know one of my personal inspirations is Nikola Tesla. In addition to being a truly masterful engineer in general, he even dabbled in robotics and remote operation, well in advance of its widespread acceptance in more modern times. I sometimes wonder just what he might have been able to do with modern computer components, considering what he was able to do with the tools and materials of his own era...
Scifi is a strong influence for me too. I think a lot of the most interesting scifi stories are the ones that explore the psychology of thinking machines in various forms; comparing and contrasting them to other sorts of beings, and assessing how near or far they may fall to human and animal ways of thinking. Perhaps it's a somewhat gratuitous choice, but one of my favourite robots is the simply-titled Robo from Chrono Trigger, as he comes to certain conclusions about his own role and importance in the game(/story)'s driving conflict, as compared to those of his(?) human companions: Roles above those of a mere tool, servant, or obliged guardian. His wasn't the most active role (and as a story, Chrono Trigger has its weaknesses), but it brought forth some interesting questions upon reflection, for me anyway.
And of course there's the scifi greats like Asimov, complex characterizations like Star Trek's Data, and more epochryphal examples in older fiction, like golems and automatons; fantasy-robots of earlier times. Metropolis (both the original German film, and the Japanese anime inspired by it and Tezuka's work) also hang in my mind, for various reasons. Do you have any personal favourites of that sort that spring to mind?
I have a lot of favourites. The Matrix had some very interesting ideas about virtual reality and 'awakening' as well as sentient machines, but the sheer nonsense of using humans as fuel made it difficult to address them to the same degree as something that might ever actually happen, and the focus on action kind of relegated those questions to passing soliloquys rather than in-depth exploration.
I think, despite its mainstream popularity and somewhat gritty take on the subject, Ghost in the Shell and its associated series best managed to capture the essence of how I see the world evolving, in terms of both flagrant depersonalization and search for purpose, as well as the clash of ideals which is already beginning to emerge. Plus, the formula of 1 philosophical episode for each action-focused episode allowed a lot more in-depth exploration without sinking into the role of animated thesis paper.
Honestly...there aren't that many 'sentient AI' characters that spark my imagination. The best exploration (for me) to try and wrap my mind around the prospect of an artificial human was actually to develop a character who is what I envision a sentient AI to be using my knowledge of the subject and role-play as that character in various settings. It was pretty interesting to see how people respond to the strange mix of detached logic, careful obedience, and innocent benevolence in an organic setting; there were those who felt the need to shelter her despite her ability to cope with the horrors of reality, there were those who simply wanted to explore her unusual take on the world, and there were those who used her unconditional positive regard as validation when they lacked it in their real lives. Very few treated her as an object or tool to be used, despite her willingness to follow instructions, even with regards to romance.
I think this experience taught me more about the possible role of a sentient robot in society than any movie or novel could.
Anyway, I think that whatever forms of robotic minds ultimately emerge from our endeavours as a species, humans will learn a tremendous amount about themselves in the process. I'm excited to see that happen, when it does.
(Oh lord, I've gone and made a gigantic text wall again. I'll figure this whole "brevity" thing out eventually... /:V )
Lol, no worries. You know, that sounds quite neat, but it's probably more than I could do anything with at this point. I've only just begun looking at neural simulation in practice, and my programming skills are very modest as of yet. I've slowly learned that diving into things headfirst, without a solid grounding to stand on beforehand, generally results in a lot of mostly-useless floundering, and not much else. X3 I really appreciate the offer, though, and perhaps I'll ask you about it again sometime, when I'm more confident in my skills. Perhaps you should post it somewhere anyway, though, if you're keen to share? May be better than letting it moulder on a lonely hard drive.
Heh, that's how I initially started programming, no skills whatsoever and I ask a friend to try and teach me something 'small'. That first project ended up being a multi-classed, multi-threaded port scanner, needless to say it took forever and could have been done a lot better, but I guess I sort-of survived.
Also, it seems like it's your lucky day, my friend and I just decided to rewrite the evolution sim using neural networks and genetic algorithms, something we are both rather excited to fool around in. It should be rather fun.
It may not seem like it from our exchanges so far, but I'm quite succinct and irreverent a lot of the time; I also don't much care for formalities. Delayed-response formats like Internet forums bring out the worst of my verbosity, in spite of myself. Plus your initial reply sort of pushed me into Debate Mode... XD ("HULK IS REFER TO HIS DETAILED LIST OF TALKING POINTS WHEN ANGRYYY"w(oD0(w )
For me, I think the value lies at least partly in the mnemonic value of a clear image or diagram, where it's easier to chunk subunits of a problem into a single visual metaphor. I agree that explanations and metaphors/examples as you describe are tremendously important in learning, though. I value both methods of capturing/relaying information; I guess I simply tend to associate deeper learning with clearer and more precise diagrammatic summaries, in my mind. Lists of detailed instructions are usually harder for me to remember. /:P (SVMs: Another thing for me to pile precariously atop my metaphorical reading pile D: )
Yeah, there's certainly the parallel aspect, but there's more...
Yes, true indeed. I've read a little bit about the computational roles of glia, and possibilities for computation by support cells and non-neurons in general within the brain, but it's admittedly not a subject I know much about. Have there been any particularly big discoveries there of late? I've not checked into that angle of inquiry for a while...
That said, I'd hope a lot of that complexity can actually be reduced to some extent in most practical applications; considering there seems little obvious value in simulating (or otherwise constructing) a brain that gets tired, or runs short on chemical fuel, or that is overheated, etc. Then again, when it comes to nature and its workings, I currently can't help but enjoy most of the interesting surprises.
For some silly reason I hadn't yet considered vectors that way, but of course that is exactly what they are. u)+c+)a ha-durrr
I think I get the general idea as you describe it, although I'm not sure I quite understand how the weightings work. If every node in a given layer connects to every node in the next layer (in this simple example), doesn't that mean that each layer is filled with identical values (assuming all nodes in a layer start with the same value)? I'm probably missing something in your explanation, though...
EDIT: ...which is probably that I'm making the really dumb assumption that all the input nodes in the FIRST layer would have the same values, and WHY WOULD THAT BE HMMM yeah no. *headdesk* I think I get it now. u);^;)a
But either way, I'll spare you from further noobish questions for the moment, until I've done more of my own reading. No sense in unnecessarily taking up your time. Thanks for the detailed explanation (you exceeded my expectations with the diagram ), it's appreciated!
I really enjoyed the Matrix, both as an action film and as a cinematic exercise (visual style, character/setting designs, etc), and certainly for its explorations of a rather unique sort of robot uprising. But yes, I think we Matrix fans all have similar grievances: X3
(spoiler'd because big)
I should refresh my memory of it, but from what I can recall offhand, I think the Animatrix supplementary release actually had quite a bit more in terms of setting exploration, with a greater focus on the machines and their coming-to alongside humans. I remember snippets of one sequence to do with human (in)tolerance of the first human-like robots in society, and also a rather dramatic, wordless sequence that I believe had to do with humans attempting to capture and reprogram (communicate with?) one of the machines. If you've not heard of this release before, you may enjoy it. It was also quite interesting visually as well (I have a strong fondness for traditional animation, often in preference to 3D, which also helps in my case).
I was somewhat reluctant to mention GitS, just because yes, it is so very very maintstream (overhyped even). XD But I too enjoy it, despite its rampant navel-gazing and occasional narrative pacing issues. I can't help but love the Tachikomas, and it often feels like one of the most well-realized near-future scifi settings I'm familiar with so far, so I certainly appreciate where you're coming from there. Honestly, I can't help but think that any technology(/ies) that so openly question (or challenge) the definition of what we are, as humans, would be quite likely to push us into some amount of psychological grittiness and angst, as we come to terms with what it means. Of course, in GitS, you see almost everything through the eyes of a military SpecOps of sorts, so of course there's a strong emphasis on the negatives. It makes me wonder what a GitS-setting story, framed in a less confrontational perspective, might look like. /:P
Anyway, it's a shame that Masamune Shirow went off on such an odd tangent, career-wise, not long after making it. His initial GitS manga are fun, for instance, but then you'll see his art and storytelling style change drastically (and I'd say much for the worse). And it seems like lately he's all but dropped off the radar. Thankfully the anime adaptations mostly avoided those particular problems. I wonder sometimes just what led to his seeming change of direction, though, as I don't really know.
Honestly...there aren't that many 'sentient AI' characters that spark my imagination. The best exploration (for me) to try and wrap my mind around the prospect of an artificial human was actually to develop a character who is what I envision a sentient AI to be...
I can't say there are a ton that really excite me either (though HAL and GLaDOS spring to mind...); yet I'll often see such a character in a fictional work, and start to think about it in more detail afterwards, and start asking more in-depth questions about it and its setting than may have come up directly in the main narrative of the work. Sometimes it's the metanarrative, more than the narrative itself, that I find engaging.
I happen to recall one good film I saw recently, though. Have you seen Moon? It's a very gritty story, and very psychological (for better or worse), about a lonely, isolated man running a mining installation on Earth's moon. But if nothing else, I suspect you'd enjoy the characterization of the robotic assistant "GERTY" (and the man's interactions with it) seen in it.
Another one perhaps worth mentioning is Robot and Frank, where an aging once-criminal man is given a nursing robot to look after him. I didn't enjoy this one as much, but even so, I think the robot's characterization was also quite well-done, within the context of the setting.
EDIT: Oh, and I just remembered. It's of a rather different bent, but I might also suggest Analogue: A Hate Story. It mostly glosses over the AI/human line, and the scifi emphasis is more human-sociological/psychological than outright focussed on AI, but it's quite well-constructed nevertheless (though perhaps afflicted with certain cliches here and there). If visual novels interest you at all, I'd encourage you to try it without looking up any walkthroughs or guides; it doesn't branch much, so it's not too overwhelming, and it has a trick or two that are worth a bit of ignorance to experience fresh on the first playthrough.
I'm rather intrigued by what you describe regarding your role-playing characterization. I imagine it's something you're not keen to discuss in detail publicly, but would you mind if I PM'd you on the subject? Or maybe there's another way I might talk with you about it? I'm currently a writer before I am a roboticist, so that sort of thinking interests me. It'd be neat to hear more about your explorations in that regard.
I would be sad if this discussion went completely private. I have been enjoying the read and mental stimulus so far. So consider me your audience up to this point. I say up to this point because I have something to add to AI characters in fiction. Cortana from halo was well done. Yes she is almost human in appearance and characterisation but that is because she was copied from the scientist responsible for the creation of the AI. You have to look to the novels for the full story though, the games focus on the action so much they don't mention anything outside what happens in the games. I can't remember which one explores AI rampancy, but as far as sci-fi goes, I'd say they're all worth the read. The space battles are exciting and use realistic distances and speeds in their descriptions, something I think is rather unusual or unheard of in other mediums. So that's my two bits.
Machine Muse, it becomes apparent why you have chosen your name. For what it's worth, I think it suits you. Your interests at least. At first I was angry at your response to my idea, but then I took it as a challenge. I'm half way through reading this thread so I guess you can read half of my idea now. I apologise for the way I presented the idea. I have reasons but no excuses for the way I came across. Since I have trouble being interesting for as long as skyeye_machine, I'll leave it at that.
Heh, that's how I initially started programming, no skills whatsoever and I ask a friend to try and teach me something 'small'. That first project ended up being a multi-classed, multi-threaded port scanner, needless to say it took forever and could have been done a lot better, but I guess I sort-of survived.
Also, it seems like it's your lucky day, my friend and I just decided to rewrite the evolution sim using neural networks and genetic algorithms, something we are both rather excited to fool around in. It should be rather fun.
Well, we've all got to start somewhere~ I find I'm not the sort that can be content with skillbuilding, though, unless I have a clear idea of the final goal, to some definitive extent. If I don't have the final product in mind, it's very difficult to figure out the meaningful steps required for a programming solution that will take me there... Regardless, it sounds like you came through with a pretty neat result in the end, s'nothing to sniff at if you ask me. /:D
Also, glad to hear your project is reviving. Genetic algorithms are also very fascinating, that does sound like it could be pretty cool. I've done a bit of reading on John Koza's work, but as with most compsci subjects, there's still a ton I don't yet understand all that well. Koza's "invention machines" are a favourite idea and inspiration of mine, though, from what I've heard of them. Last I read on the subject, it seemed like they'd even produced some patentable inventions, including a very potent little radio antenna, although I've heard little about such developments lately. I should go check again...
Anyway, as for what you're doing, where are you and your friend planning to take this project exactly? Is it more of a general framework for genetic algorithms/neural sims, or are you trying to engineer something specific using such approaches?
The end goal is a two dimensional evolution simulator (with only one 'species' of bot and one food source). The project is a sort-of precursor to another project which might utilize a neural network to do all sorts of nifty things. (and by precursor, I mean a test to see if we can successfully make a functioning, evolving neural network)
The idea is to link a mass of inputs and outputs into a randomly generated tree structure and then utilize genetic algorithms to slowly morph it into a whole process tree that is ideal for survival in the generated world.
So, when I install the addon pack for 1.5.1 I get this:
Minecraft has crashed!
----------------------
Minecraft has stopped running because it encountered a problem; Failed to start game
A full error report has been saved to C:\Users\Isaac\AppData\Roaming\.minecraft\crash-reports\crash-2013-04-13_12.28.55-client.txt - Please include a copy of that file (Not this screen!) if you report this crash to anyone; without it, they will not be able to help fix the crash
--- BEGIN ERROR REPORT 7678dec6 --------
Full report at:
C:\Users\Isaac\AppData\Roaming\.minecraft\crash-reports\crash-2013-04-13_12.28.55-client.txt
Please show that file to Mojang, NOT just this screen!
cpw.mods.fml.common.LoaderException: java.lang.NoClassDefFoundError: andrew/powersuits/modules/OmniWrenchModule
at cpw.mods.fml.common.LoadController.transition(LoadController.java:142)
at cpw.mods.fml.common.Loader.initializeMods(Loader.java:690)
at cpw.mods.fml.client.FMLClientHandler.finishMinecraftLoading(FMLClientHandler.java:206)
at net.minecraft.client.Minecraft.func_71384_a(Minecraft.java:444)
at net.minecraft.client.MinecraftAppletImpl.func_71384_a(SourceFile:56)
at net.minecraft.client.Minecraft.run(Minecraft.java:729)
at java.lang.Thread.run(Unknown Source)
Caused by: java.lang.NoClassDefFoundError: andrew/powersuits/modules/OmniWrenchModule
at andrew.powersuits.common.AddonConfig.loadPowerModules(AddonConfig.java:36)
at andrew.powersuits.common.ModularPowersuitsAddons.load(ModularPowersuitsAddons.java:38)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at cpw.mods.fml.common.FMLModContainer.handleModStateEvent(FMLModContainer.java:515)
at sun.reflect.GeneratedMethodAccessor5.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at com.google.common.eventbus.EventHandler.handleEvent(EventHandler.java:74)
at com.google.common.eventbus.SynchronizedEventHandler.handleEvent(SynchronizedEventHandler.java:45)
at com.google.common.eventbus.EventBus.dispatch(EventBus.java:314)
at com.google.common.eventbus.EventBus.dispatchQueuedEvents(EventBus.java:296)
at com.google.common.eventbus.EventBus.post(EventBus.java:267)
at cpw.mods.fml.common.LoadController.propogateStateMessage(LoadController.java:165)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at com.google.common.eventbus.EventHandler.handleEvent(EventHandler.java:74)
at com.google.common.eventbus.SynchronizedEventHandler.handleEvent(SynchronizedEventHandler.java:45)
at com.google.common.eventbus.EventBus.dispatch(EventBus.java:314)
at com.google.common.eventbus.EventBus.dispatchQueuedEvents(EventBus.java:296)
at com.google.common.eventbus.EventBus.post(EventBus.java:267)
at cpw.mods.fml.common.LoadController.distributeStateMessage(LoadController.java:98)
at cpw.mods.fml.common.Loader.initializeMods(Loader.java:689)
... 5 more
Caused by: java.lang.ClassNotFoundException: andrew.powersuits.modules.OmniWrenchModule
at cpw.mods.fml.relauncher.RelaunchClassLoader.findClass(RelaunchClassLoader.java:211)
at java.lang.ClassLoader.loadClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
... 32 more
Caused by: java.lang.NoClassDefFoundError: buildcraft/api/tools/IToolWrench
at java.lang.ClassLoader.defineClass1(Native Method)
at java.lang.ClassLoader.defineClass(Unknown Source)
at java.security.SecureClassLoader.defineClass(Unknown Source)
at cpw.mods.fml.relauncher.RelaunchClassLoader.findClass(RelaunchClassLoader.java:200)
... 34 more
Caused by: java.lang.ClassNotFoundException: buildcraft.api.tools.IToolWrench
at cpw.mods.fml.relauncher.RelaunchClassLoader.findClass(RelaunchClassLoader.java:113)
at java.lang.ClassLoader.loadClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
... 38 more
--- END ERROR REPORT 9afb3e83 ----------
The end goal is a two dimensional evolution simulator (with only one 'species' of bot and one food source). The project is a sort-of precursor to another project which might utilize a neural network to do all sorts of nifty things. (and by precursor, I mean a test to see if we can successfully make a functioning, evolving neural network)
The idea is to link a mass of inputs and outputs into a randomly generated tree structure and then utilize genetic algorithms to slowly morph it into a whole process tree that is ideal for survival in the generated world.
At the risk of being discouraging, I'm surprised you've not heard(?) of Polyworld before now, considering it's basically exactly what you describe. Perhaps you could learn a thing or two from their implementation:
That said, I'd certainly encourage you to try and develop your ideas; there's always room for more exploration. I think the evolutionary/genetic approach is a very interesting one for a number of reasons; and in a more fantastic vein, I can't help but hope something not unlike the Darwinians (from Introversion's Darwinia) will eventually come to exist in the real world, if only because I'm curious as to what is possible. Plus it'd probably be quite a fascinating experience, to interact with such entities. It was fun enough in the game.
MachineMuse
1) Please add a Lappack/Batpack/advanced Lappack module
2) Please add a creative battery module that only appears in the table when in Creative/ (Neis Creative+ if possible) so you have infinite amounts of energy always
so like 10000MJ/10000MJ Forever
3) I will add more as i think
Rollback Post to RevisionRollBack
"Schizophrenia is a culturally bound phenomenon. Its pattern of expression is filtered through the cultural substrate in which its symptoms develop,in technological Societies, this manifests as delusions of surveillance and a belief that advanced technology is deployed against you, usually with some vague unseen "other" out to get you."
MachineMuse
1) Please add a Lappack/Batpack/advanced Lappack module
2) Please add a creative battery module that only appears in the table when in Creative/ (Neis Creative+ if possible) so you have infinite amounts of energy always
so like 10000MJ/10000MJ Forever
3) I will add more as i think
What would adding a lappack/batpack module even do? If you mean "make electric tools draw from the armor before drawing from themselves" you should say that.
If you have UE you can get infinite power in creative with the use of an infinite battery. I'm really suprised that idea worked, but it did, so if you need limitless creative power: there ya go. As in just, have an infinite battery in your inventory. Your power gauge turns into an infinity. Its totes boss.
I'll get to you in a sec. I just had to finish this. Build #329 is now recommended!
-GUI now shows what modules are installed, as well as which ones are disabled in config.
-Module list can now be scrolled - small blue arrows will indicate when there is more on a page, and you can click the arrows or use the mousewheel to scroll. This mostly won't be noticed unless Andrew2448's addons is installed (which have had quite a few new additions!).
-Various bugfixes (item names, API mismatches)
-Custom models are go!
As promised, the visor and the crystals can be coloured separately from the main armor bits, and they glow in the dark! Extra bits and animations still to come, but this sets the groundwork for a variety of other .obj armor models if people want to make them.
Lol. That approach sounds like it has an interesting story or two behind it...and I can't say I wasn't somewhat amused by your openly blunt response, admittedly. Still, I hope that perhaps you'll be slightly more forgiving when you next set your sights upon me. I think I'm one of the more decent sorts of characters bouncing around the internet, all in all.
I have a surprisingly high tolerance for not-easy-to-follow. ;3 Linear algebra, though, that's neat; I keep seeing references to its importance in the machine-learning field. I've not got that far yet; mostly I've been learning about the basics of calculus and integration in relation to physics studies I've been pursuing lately. No doubt I'm a lightweight at the number-crunching compared to you at this point. Do you find it to be a difficult subject? I understand computers do most of the heavy numerical lifting, but of course you have to formulate everything right or they spit out garbage, as always... X3
My math background isn't that strong, but I enjoy learning about the numerical side of nature nonetheless when I understand the physical ideas behind the numbers. I'm still figuring out what aspects of robotics and machine learning most attract me, but I've always loved robots of almost all sorts, and I know I find things like the Blue Brain project, and Kwabena Boahen's Neurogrid work, very fascinating. Do you have any special influences in that regard?
I'm especially keen to understand more about how neurons operate computationally, and the ways in which they can be reliably modeled in that respect, though I'm a ways off from any meaningful competence in that regard. Brains are just the most interesting piles of goop...it's hard not to get ahead of myself in my excitement, though, when thinking about the ways in which biological and technological solutions for observing and interacting with the world increasingly overlap.
Anyway, I'm curious now; where do you see yourself headed in the future, with regards to your current studies? Are there any memorable experiences you've had that particularly cemented those interests for you, as compared to other pursuits? (I hope I'm not being too presumptuous with my questions~)
Oh and @MachineMuse, are animations allowed? (If so, would I make them as a single file that is about multipleof16*16 pixels... or would I have to make separate
files?)
#1 in Crazy-ness
Guess how I feel.
Have you considered the alternative of just...not posting?
Hehe, that's delightful! One of my major goals was to have different options for different playstyles, so I love hearing of different ways people specialize. Hopefully as MPS gets more refined and closer to something that could be considered a 1.0 release, the options for diversity will increase.
oh, so now you're going to encourage me?
Yeah, the math is kinda the hard part for me, too - it is actually very simple stuff in practice but wrapping my head around what it all means is proving a bit of a chore. But it's important, because translating everything into low-level operations means it can be optimized at that level, often in a way that makes the difference between time-complexity classes.
But when it comes to simulations? Computers and brains are so fundamentally different from each other that in my experience it's more effective to focus on things that each does well, and on interfacing them in ways that effectively bridge the gaps in their capabilities.
The simplest non-trivial brain-like data structure is the feed-forward neural network, so that's something to look into if you're just spreading your wings in this subject. The trick is, each 'layer' of neurons is actually just a matrix (representing a linear system) with a non-linear transformation function on the output, which is what makes it so you can start to do some really fancy math with it. And from there, you can start to develop it in ways that take advantage of the things a computer is good at, as opposed to a human.
I'm really more inspired by scifi on the topic and by the neuropsych courses I've taken which have shown me that the common conception 'We don't really understand how the brain works' is pretty far from the truth. I don't have any particularly strong focuses; the whole field is just so fascinating to me that every little development is exciting. I would more point to specific people than experiences, though, which I am not thrilled about doing on a public forum. I will at least mention my late grandmother, who sparked my interest in design and technology very early with her legos and engineer toys (Jacob's ladder, Newton's cradle, ring-and-string puzzles, etc.)
Yeah, animations in Minecraft are just tiled vertically now. Check out the lava texture in minecraft.jar if you need a demo.
Here is that updated link for the Mod Spotlight for 0.3.2-199
Thanks!
Glad you like it and thanks for the support!
Ok, I put it on my site. Thanks!
Only a little. ;D
I value forthrightness in discussion (though it can be disgruntling at times), and sometimes a measure of open controversy is a refreshing alternative to stifled politenesses. The world has enough legalese in it as it is. /:P
Math is always very simple, in a way (or so I tell myself at the times I'm most stumped by it ). I mean, the expressions of a given function, etc, can be very complex and nuanced (not that I'm all that much of a mathematician), but in the end it all comes down to relatively simple logical foundations, stacked on top of each other in varyingly sophisticated - but unambiguous - ways, so far as I understand. I suppose it's more that human active memory chunking limitations are to blame. I know I run into them often enough. X3 I read some of Spivak's text on calculus and analysis, and while a lot of it is beyond my grasp, it definitely reinforced the inherrent logical rigour of the process in a rather adept way that I still try and hold on to in my other studies.
Personally, I usually find it's mostly a matter of grasping the deeper context, when it comes to having a sense of the numbers and operators. I think I have a rather visual bent to my mind; trigonometry is one of the easier things for me to grasp intuitively because it's easier to visualize the geometry in ways that clearly illustrate the core relationships. Finding those clarifying images is for me one of the more gratifying aspects of the learning process.
I don't know much about optimization concerns, seeing as I've never developed anything to the point where it matters, so I lack that insight. I'm aware of how important it can be in practice, however; one of the more memorable examples I know of is Quake's Fast InvSqrt() function in the days of yore. Could you elaborate on what you mean regarding "time-complexity classes"? Also, as for low-level operations, how low do most programmers typically go these days? Is there much place for assembly languages anymore, or would that go beyond what's typically expected in most development scenarios?
Yes, I'm reasonably aware of how inherrently different brains and classical CPUs are, and I agree, a hybrid approach seems the best method all around. When I talked about convergence before, I was referring more to our growing capacity to bridge these two very distinctive approaches in a meaningful way. The Blue Brain project is sort of insane when you consider the hardware they're running it on; they're trying to emulate a massively parallel system with an inherrently sequential approach, and that's the main reason it's as expensive as it is. It's an inefficient realization, though of course it makes sense in context (economy, available tech, familiarity, etc). I'm very excited to see how memristors start to change things as they're developed and introduced commercially over the next few decades, since they'll seem to encourage some more innovative, inherrently parallel archetectures.
Regarding feed-forward neural networks; that makes sense, thanks for the insight. I've been reading a bit on how individual neurons can be modeled mathematically (the "leaky-integrate-and-fire" model, for example), but have not considered them in that broader, interlinked context yet. Out of curiousity, am I right to presume that in a simple feed-forward network as you describe, each layer of N neurons in a hierarchy of layers can be treated as a 1-by-N (horizontal) matrice (that is, can a single neuron in a layer be destribed by a single function, or is more than one function (columns) needed)? Or would that depend on the neural model used? (Either way, more to learn on my part~)
Yes, it's hard to be selective in excitement, isn't it, with everything going on~ Especially in how each advance in a related field enhances the achievements of the others. And it's pretty incredible how far our understandings of brains have come. I was reading not long ago about a monkey that has been trained to control a third robotic arm through neural activity alone (back in 2010!); and of course there are exciting things happening in prosthetics more generally as well. And then of course there's all the buzz over drones, both autonomous and not. For better or worse, we live in interesting times.
Also, I get your concerns about posting that sort of stuff publicly, no problem. I simply enjoy talking about these topics, so please pardon me if I'm a bit overbearing. Regardless, I think it's great that your grandmother was a supportive influence as she was; that sort of rapport is important as a kid, when you're figuring out where to go in life. (I've still got all my own old lego buried somewhere...) I know one of my personal inspirations is Nikola Tesla. In addition to being a truly masterful engineer in general, he even dabbled in robotics and remote operation, well in advance of its widespread acceptance in more modern times. I sometimes wonder just what he might have been able to do with modern computer components, considering what he was able to do with the tools and materials of his own era...
Scifi is a strong influence for me too. I think a lot of the most interesting scifi stories are the ones that explore the psychology of thinking machines in various forms; comparing and contrasting them to other sorts of beings, and assessing how near or far they may fall to human and animal ways of thinking. Perhaps it's a somewhat gratuitous choice, but one of my favourite robots is the simply-titled Robo from Chrono Trigger, as he comes to certain conclusions about his own role and importance in the game(/story)'s driving conflict, as compared to those of his(?) human companions: Roles above those of a mere tool, servant, or obliged guardian. His wasn't the most active role (and as a story, Chrono Trigger has its weaknesses), but it brought forth some interesting questions upon reflection, for me anyway.
And of course there's the scifi greats like Asimov, complex characterizations like Star Trek's Data, and more epochryphal examples in older fiction, like golems and automatons; fantasy-robots of earlier times. Metropolis (both the original German film, and the Japanese anime inspired by it and Tezuka's work) also hang in my mind, for various reasons. Do you have any personal favourites of that sort that spring to mind?
Anyway, I think that whatever forms of robotic minds ultimately emerge from our endeavours as a species, humans will learn a tremendous amount about themselves in the process. I'm excited to see that happen, when it does.
(Oh lord, I've gone and made a gigantic text wall again. One day I'll figure this whole "brevity" thing out, and my replies to things will never exceed 250 words, and the internet will ring with bright rejoicing, and maybe I'll even get a firm, brisk handshake from the Singularity... =c= )
Lol, no worries. You know, that sounds quite neat, but it's probably more than I could do anything with at this point. I've only just begun looking at neural simulation in practice, and my programming skills are very modest as of yet. I've slowly learned that diving into things headfirst, without a solid grounding to stand on beforehand, generally results in a lot of mostly-useless floundering, and not much else. X3 I really appreciate the offer, though, and perhaps I'll ask you about it again sometime, when I'm more confident in my skills. Perhaps you should post it somewhere anyway, though, if you're keen to share? May be better than letting it moulder on a lonely hard drive. /
I'm surprised you understand, tbh! Oh well.
Hm, visual analogies sometimes help but I usually find sequential/verbal explanations and metaphors more useful. A visual analogy is meaningless without an understanding of what each part is trying to do. Although for SVMs it really does help...
For most programmers, it hardly matters at all, because you aren't dealing with time-critical code. Even in graphics programming, a loss of 2-3 fps on an average computer is hardly noticeable. But there is still a niche for it, because in some industries the amount of information you can process is directly correlated to the amount of money you can make, and even when not, there is still a significant cost savings if, for example, someone can make a few code changes at some bottleneck and avoid needing another server rack.
Time-complexity classes are a different issue, and one that's more related to classical computer science. It's really a pretty broad topic so I'll direct you to wikipedia, since this article is pretty well-written imo:
http://en.wikipedia....Time_complexity
I see. Yeah, there's certainly the parallel aspect, but there's more - for a really accurate brain simulation, you need to account for varying resource availability and environmental conditions in different regions of the brain, as well as spatial proximity and even the role of support cells (glia). There's probably more still.
As for the neural network: remember that each neuron (in the simplest model) is connected to each neuron in the following layer. So the actual state of the neuron can be a 1-by-N vector (a vector is just a matrix with a single row or column) representing the input or stimulation that each of them is receiving, but the actual data encoded in the neurons is M-by-N (where M is the number of neurons in the following layer). In other words: you can think of the rows as the 'weights' that a given 2nd-level neuron takes from each of the 1st-level neurons' outputs. So each output neuron has a value equal to a different weighted sum of the input neurons' values. Then you apply some transformation to it, probably a sigmoid:
To get a value between 0 and 1, representing the probability that that neuron is going to fire.
I have a lot of favourites. The Matrix had some very interesting ideas about virtual reality and 'awakening' as well as sentient machines, but the sheer nonsense of using humans as fuel made it difficult to address them to the same degree as something that might ever actually happen, and the focus on action kind of relegated those questions to passing soliloquys rather than in-depth exploration.
I think, despite its mainstream popularity and somewhat gritty take on the subject, Ghost in the Shell and its associated series best managed to capture the essence of how I see the world evolving, in terms of both flagrant depersonalization and search for purpose, as well as the clash of ideals which is already beginning to emerge. Plus, the formula of 1 philosophical episode for each action-focused episode allowed a lot more in-depth exploration without sinking into the role of animated thesis paper.
Honestly...there aren't that many 'sentient AI' characters that spark my imagination. The best exploration (for me) to try and wrap my mind around the prospect of an artificial human was actually to develop a character who is what I envision a sentient AI to be using my knowledge of the subject and role-play as that character in various settings. It was pretty interesting to see how people respond to the strange mix of detached logic, careful obedience, and innocent benevolence in an organic setting; there were those who felt the need to shelter her despite her ability to cope with the horrors of reality, there were those who simply wanted to explore her unusual take on the world, and there were those who used her unconditional positive regard as validation when they lacked it in their real lives. Very few treated her as an object or tool to be used, despite her willingness to follow instructions, even with regards to romance.
I think this experience taught me more about the possible role of a sentient robot in society than any movie or novel could.
For sure. I do the same thing >.<
Heh, that's how I initially started programming, no skills whatsoever and I ask a friend to try and teach me something 'small'. That first project ended up being a multi-classed, multi-threaded port scanner, needless to say it took forever and could have been done a lot better, but I guess I sort-of survived.
Also, it seems like it's your lucky day, my friend and I just decided to rewrite the evolution sim using neural networks and genetic algorithms, something we are both rather excited to fool around in. It should be rather fun.
#1 in Crazy-ness
It relies heavily on forge hooks to be compatible with other mods.
It may not seem like it from our exchanges so far, but I'm quite succinct and irreverent a lot of the time; I also don't much care for formalities. Delayed-response formats like Internet forums bring out the worst of my verbosity, in spite of myself. Plus your initial reply sort of pushed me into Debate Mode... XD ("HULK IS REFER TO HIS DETAILED LIST OF TALKING POINTS WHEN ANGRYYY" w(oD0(w )
For me, I think the value lies at least partly in the mnemonic value of a clear image or diagram, where it's easier to chunk subunits of a problem into a single visual metaphor. I agree that explanations and metaphors/examples as you describe are tremendously important in learning, though. I value both methods of capturing/relaying information; I guess I simply tend to associate deeper learning with clearer and more precise diagrammatic summaries, in my mind. Lists of detailed instructions are usually harder for me to remember. /:P (SVMs: Another thing for me to pile precariously atop my metaphorical reading pile D: )
Makes sense~
I can see I'll need to read through this in some depth before I'll have much worth saying in reply. Thanks for the link, it looks interesting.
Yes, true indeed. I've read a little bit about the computational roles of glia, and possibilities for computation by support cells and non-neurons in general within the brain, but it's admittedly not a subject I know much about. Have there been any particularly big discoveries there of late? I've not checked into that angle of inquiry for a while...
That said, I'd hope a lot of that complexity can actually be reduced to some extent in most practical applications; considering there seems little obvious value in simulating (or otherwise constructing) a brain that gets tired, or runs short on chemical fuel, or that is overheated, etc. Then again, when it comes to nature and its workings, I currently can't help but enjoy most of the interesting surprises.
For some silly reason I hadn't yet considered vectors that way, but of course that is exactly what they are. u)+c+)a ha-durrr
I think I get the general idea as you describe it, although I'm not sure I quite understand how the weightings work. If every node in a given layer connects to every node in the next layer (in this simple example), doesn't that mean that each layer is filled with identical values (assuming all nodes in a layer start with the same value)? I'm probably missing something in your explanation, though...
EDIT: ...which is probably that I'm making the really dumb assumption that all the input nodes in the FIRST layer would have the same values, and WHY WOULD THAT BE HMMM yeah no. *headdesk* I think I get it now. u);^;)a
But either way, I'll spare you from further noobish questions for the moment, until I've done more of my own reading. No sense in unnecessarily taking up your time. Thanks for the detailed explanation (you exceeded my expectations with the diagram ), it's appreciated!
I really enjoyed the Matrix, both as an action film and as a cinematic exercise (visual style, character/setting designs, etc), and certainly for its explorations of a rather unique sort of robot uprising. But yes, I think we Matrix fans all have similar grievances: X3
(spoiler'd because big)
I should refresh my memory of it, but from what I can recall offhand, I think the Animatrix supplementary release actually had quite a bit more in terms of setting exploration, with a greater focus on the machines and their coming-to alongside humans. I remember snippets of one sequence to do with human (in)tolerance of the first human-like robots in society, and also a rather dramatic, wordless sequence that I believe had to do with humans attempting to capture and reprogram (communicate with?) one of the machines. If you've not heard of this release before, you may enjoy it. It was also quite interesting visually as well (I have a strong fondness for traditional animation, often in preference to 3D, which also helps in my case).
I was somewhat reluctant to mention GitS, just because yes, it is so very very maintstream (overhyped even). XD But I too enjoy it, despite its rampant navel-gazing and occasional narrative pacing issues. I can't help but love the Tachikomas, and it often feels like one of the most well-realized near-future scifi settings I'm familiar with so far, so I certainly appreciate where you're coming from there. Honestly, I can't help but think that any technology(/ies) that so openly question (or challenge) the definition of what we are, as humans, would be quite likely to push us into some amount of psychological grittiness and angst, as we come to terms with what it means. Of course, in GitS, you see almost everything through the eyes of a military SpecOps of sorts, so of course there's a strong emphasis on the negatives. It makes me wonder what a GitS-setting story, framed in a less confrontational perspective, might look like. /:P
Anyway, it's a shame that Masamune Shirow went off on such an odd tangent, career-wise, not long after making it. His initial GitS manga are fun, for instance, but then you'll see his art and storytelling style change drastically (and I'd say much for the worse). And it seems like lately he's all but dropped off the radar. Thankfully the anime adaptations mostly avoided those particular problems. I wonder sometimes just what led to his seeming change of direction, though, as I don't really know.
I can't say there are a ton that really excite me either (though HAL and GLaDOS spring to mind...); yet I'll often see such a character in a fictional work, and start to think about it in more detail afterwards, and start asking more in-depth questions about it and its setting than may have come up directly in the main narrative of the work. Sometimes it's the metanarrative, more than the narrative itself, that I find engaging.
I happen to recall one good film I saw recently, though. Have you seen Moon? It's a very gritty story, and very psychological (for better or worse), about a lonely, isolated man running a mining installation on Earth's moon. But if nothing else, I suspect you'd enjoy the characterization of the robotic assistant "GERTY" (and the man's interactions with it) seen in it.
Another one perhaps worth mentioning is Robot and Frank, where an aging once-criminal man is given a nursing robot to look after him. I didn't enjoy this one as much, but even so, I think the robot's characterization was also quite well-done, within the context of the setting.
EDIT: Oh, and I just remembered. It's of a rather different bent, but I might also suggest Analogue: A Hate Story. It mostly glosses over the AI/human line, and the scifi emphasis is more human-sociological/psychological than outright focussed on AI, but it's quite well-constructed nevertheless (though perhaps afflicted with certain cliches here and there). If visual novels interest you at all, I'd encourage you to try it without looking up any walkthroughs or guides; it doesn't branch much, so it's not too overwhelming, and it has a trick or two that are worth a bit of ignorance to experience fresh on the first playthrough.
I'm rather intrigued by what you describe regarding your role-playing characterization. I imagine it's something you're not keen to discuss in detail publicly, but would you mind if I PM'd you on the subject? Or maybe there's another way I might talk with you about it? I'm currently a writer before I am a roboticist, so that sort of thinking interests me. It'd be neat to hear more about your explorations in that regard.
Machine Muse, it becomes apparent why you have chosen your name. For what it's worth, I think it suits you. Your interests at least. At first I was angry at your response to my idea, but then I took it as a challenge. I'm half way through reading this thread so I guess you can read half of my idea now. I apologise for the way I presented the idea. I have reasons but no excuses for the way I came across. Since I have trouble being interesting for as long as skyeye_machine, I'll leave it at that.
Well, we've all got to start somewhere~ I find I'm not the sort that can be content with skillbuilding, though, unless I have a clear idea of the final goal, to some definitive extent. If I don't have the final product in mind, it's very difficult to figure out the meaningful steps required for a programming solution that will take me there... Regardless, it sounds like you came through with a pretty neat result in the end, s'nothing to sniff at if you ask me. /:D
Also, glad to hear your project is reviving. Genetic algorithms are also very fascinating, that does sound like it could be pretty cool. I've done a bit of reading on John Koza's work, but as with most compsci subjects, there's still a ton I don't yet understand all that well. Koza's "invention machines" are a favourite idea and inspiration of mine, though, from what I've heard of them. Last I read on the subject, it seemed like they'd even produced some patentable inventions, including a very potent little radio antenna, although I've heard little about such developments lately. I should go check again...
Anyway, as for what you're doing, where are you and your friend planning to take this project exactly? Is it more of a general framework for genetic algorithms/neural sims, or are you trying to engineer something specific using such approaches?
The idea is to link a mass of inputs and outputs into a randomly generated tree structure and then utilize genetic algorithms to slowly morph it into a whole process tree that is ideal for survival in the generated world.
#1 in Crazy-ness
Minecraft has crashed!
----------------------
Minecraft has stopped running because it encountered a problem; Failed to start game
A full error report has been saved to C:\Users\Isaac\AppData\Roaming\.minecraft\crash-reports\crash-2013-04-13_12.28.55-client.txt - Please include a copy of that file (Not this screen!) if you report this crash to anyone; without it, they will not be able to help fix the crash
--- BEGIN ERROR REPORT 7678dec6 --------
Full report at:
C:\Users\Isaac\AppData\Roaming\.minecraft\crash-reports\crash-2013-04-13_12.28.55-client.txt
Please show that file to Mojang, NOT just this screen!
Generated 4/13/13 12:28 PM
-- System Details --
Details:
Minecraft Version: 1.5.1
Operating System: Windows 7 (amd64) version 6.1
Java Version: 1.7.0_04, Oracle Corporation
Java VM Version: Java HotSpot(TM) 64-Bit Server VM (mixed mode), Oracle Corporation
Memory: 308575176 bytes (294 MB) / 612630528 bytes (584 MB) up to 954466304 bytes (910 MB)
JVM Flags: 2 total; -Xms512m -Xmx1024m
AABB Pool Size: 0 (0 bytes; 0 MB) allocated, 0 (0 bytes; 0 MB) used
Suspicious classes: FML and Forge are installed
IntCache: cache: 0, tcache: 0, allocated: 0, tallocated: 0
FML: MCP v7.44 FML v5.1.8.611 Minecraft Forge 7.7.1.611 29 mods loaded, 29 mods active
mcp [Minecraft Coder Pack] (minecraft.jar) Unloaded->Constructed->Pre-initialized->Initialized
FML [Forge Mod Loader] (coremods) Unloaded->Constructed->Pre-initialized->Initialized
Forge [Minecraft Forge] (coremods) Unloaded->Constructed->Pre-initialized->Initialized
CodeChickenCore [CodeChicken Core] (coremods) Unloaded->Constructed->Pre-initialized->Initialized
NotEnoughItems [Not Enough Items] (coremods) Unloaded->Constructed->Pre-initialized->Initialized
HarkenScythe_Core [HarkenScythe] (1.5.1HarkenScythe_v2.0.6Forge.zip) Unloaded->Constructed->Pre-initialized->Initialized
iChunUtil [iChunUtil] (iChunUtil1.0.0.zip) Unloaded->Constructed->Pre-initialized->Initialized
Grinder [Grinder] (AttachableGrinder1.0.0.zip) Unloaded->Constructed->Pre-initialized->Initialized
Atum [Atum] (Atum 0.1.zip) Unloaded->Constructed->Pre-initialized->Initialized
BasicComponents [Basic Components] (BasicComponents_v1.3.1.70.jar) Unloaded->Constructed->Pre-initialized->Initialized
BiblioCraft [BiblioCraft] (BiblioCraft[v1.1.5].zip) Unloaded->Constructed->Pre-initialized->Initialized
ChickenChunks [ChickenChunks] (ChickenChunks 1.3.2.3.jar) Unloaded->Constructed->Pre-initialized->Initialized
Chisel [Chisel] (chisel-1.5.1-1.1.0.jar) Unloaded->Constructed->Pre-initialized->Initialized
CustomStuff2 [Custom Stuff 2] (customStuff2_0.9.11_Universal.zip) Unloaded->Constructed->Pre-initialized->Initialized
EnderStorage [EnderStorage] (EnderStorage 1.4.2.4.jar) Unloaded->Constructed->Pre-initialized->Initialized
GraviGun [GraviGun] (GravityGun1.5.0v2.zip) Unloaded->Constructed->Pre-initialized->Initialized
Hats [Hats] (Hats1.0.0.zip) Unloaded->Constructed->Pre-initialized->Initialized
iChun_Core [iChun_Core] (iChunUtil1.0.0.zip) Unloaded->Constructed->Pre-initialized->Initialized
IronChest [Iron Chest] (ironchest-universal-1.5.1-5.2.2.350.zip) Unloaded->Constructed->Pre-initialized->Initialized
mmmPowersuits [MachineMuse's Modular Powersuits] (ModularPowersuits-0.4.1-292.jar) Unloaded->Constructed->Pre-initialized->Initialized
PowersuitAddons [Andrew2448's Modular Powersuits Addon] (MPSA-0.2.1-32_MPS-292.jar) Unloaded->Constructed->Pre-initialized->Errored
MutantCreatures [Mutant Creatures] (Mutant Creatures v1.3.3 mc1.5.1.zip) Unloaded->Constructed->Pre-initialized->Initialized
Natura [Natura] (Natura_1.5.1_2.0.9.jar) Unloaded->Constructed->Pre-initialized->Initialized
PortalGun [PortalGun] (PortalGun1.5.0v2.zip) Unloaded->Constructed->Pre-initialized->Initialized
RopesPlus [Ropes+] (RopePlus_1.5.1.zip) Unloaded->Constructed->Pre-initialized->Initialized
StevesCarts [Steve's Carts] (StevesCarts2.0.0.a96.zip) Unloaded->Constructed->Pre-initialized->Initialized
TConstruct [Tinkers' Construct] (TConstruct_1.5.1_1.2.28.jar) Unloaded->Constructed->Pre-initialized->Initialized
tukmc_Vz [TukMC] (TukMC_latest.zip) Unloaded->Constructed->Pre-initialized->Initialized
TwilightForest [The Twilight Forest] (twilightforest-1.17.1.zip) Unloaded->Constructed->Pre-initialized->Initialized
LWJGL: 2.4.2
OpenGL: GeForce GT 520/PCIe/SSE2 GL version 4.3.0, NVIDIA Corporation
Is Modded: Definitely; Client brand changed to 'forge,fml'
Type: Client (map_client.txt)
Texture Pack: Default
Profiler Position: N/A (disabled)
Vec3 Pool Size: ~~ERROR~~ NullPointerException: null
cpw.mods.fml.common.LoaderException: java.lang.NoClassDefFoundError: andrew/powersuits/modules/OmniWrenchModule
at cpw.mods.fml.common.LoadController.transition(LoadController.java:142)
at cpw.mods.fml.common.Loader.initializeMods(Loader.java:690)
at cpw.mods.fml.client.FMLClientHandler.finishMinecraftLoading(FMLClientHandler.java:206)
at net.minecraft.client.Minecraft.func_71384_a(Minecraft.java:444)
at net.minecraft.client.MinecraftAppletImpl.func_71384_a(SourceFile:56)
at net.minecraft.client.Minecraft.run(Minecraft.java:729)
at java.lang.Thread.run(Unknown Source)
Caused by: java.lang.NoClassDefFoundError: andrew/powersuits/modules/OmniWrenchModule
at andrew.powersuits.common.AddonConfig.loadPowerModules(AddonConfig.java:36)
at andrew.powersuits.common.ModularPowersuitsAddons.load(ModularPowersuitsAddons.java:38)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at cpw.mods.fml.common.FMLModContainer.handleModStateEvent(FMLModContainer.java:515)
at sun.reflect.GeneratedMethodAccessor5.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at com.google.common.eventbus.EventHandler.handleEvent(EventHandler.java:74)
at com.google.common.eventbus.SynchronizedEventHandler.handleEvent(SynchronizedEventHandler.java:45)
at com.google.common.eventbus.EventBus.dispatch(EventBus.java:314)
at com.google.common.eventbus.EventBus.dispatchQueuedEvents(EventBus.java:296)
at com.google.common.eventbus.EventBus.post(EventBus.java:267)
at cpw.mods.fml.common.LoadController.propogateStateMessage(LoadController.java:165)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)
at java.lang.reflect.Method.invoke(Unknown Source)
at com.google.common.eventbus.EventHandler.handleEvent(EventHandler.java:74)
at com.google.common.eventbus.SynchronizedEventHandler.handleEvent(SynchronizedEventHandler.java:45)
at com.google.common.eventbus.EventBus.dispatch(EventBus.java:314)
at com.google.common.eventbus.EventBus.dispatchQueuedEvents(EventBus.java:296)
at com.google.common.eventbus.EventBus.post(EventBus.java:267)
at cpw.mods.fml.common.LoadController.distributeStateMessage(LoadController.java:98)
at cpw.mods.fml.common.Loader.initializeMods(Loader.java:689)
... 5 more
Caused by: java.lang.ClassNotFoundException: andrew.powersuits.modules.OmniWrenchModule
at cpw.mods.fml.relauncher.RelaunchClassLoader.findClass(RelaunchClassLoader.java:211)
at java.lang.ClassLoader.loadClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
... 32 more
Caused by: java.lang.NoClassDefFoundError: buildcraft/api/tools/IToolWrench
at java.lang.ClassLoader.defineClass1(Native Method)
at java.lang.ClassLoader.defineClass(Unknown Source)
at java.security.SecureClassLoader.defineClass(Unknown Source)
at cpw.mods.fml.relauncher.RelaunchClassLoader.findClass(RelaunchClassLoader.java:200)
... 34 more
Caused by: java.lang.ClassNotFoundException: buildcraft.api.tools.IToolWrench
at cpw.mods.fml.relauncher.RelaunchClassLoader.findClass(RelaunchClassLoader.java:113)
at java.lang.ClassLoader.loadClass(Unknown Source)
at java.lang.ClassLoader.loadClass(Unknown Source)
... 38 more
--- END ERROR REPORT 9afb3e83 ----------
At the risk of being discouraging, I'm surprised you've not heard(?) of Polyworld before now, considering it's basically exactly what you describe. Perhaps you could learn a thing or two from their implementation:
http://www.google.ca...rch?q=polyworld
That said, I'd certainly encourage you to try and develop your ideas; there's always room for more exploration. I think the evolutionary/genetic approach is a very interesting one for a number of reasons; and in a more fantastic vein, I can't help but hope something not unlike the Darwinians (from Introversion's Darwinia) will eventually come to exist in the real world, if only because I'm curious as to what is possible. Plus it'd probably be quite a fascinating experience, to interact with such entities. It was fun enough in the game.
1) Please add a Lappack/Batpack/advanced Lappack module
2) Please add a creative battery module that only appears in the table when in Creative/ (Neis Creative+ if possible) so you have infinite amounts of energy always
so like 10000MJ/10000MJ Forever
3) I will add more as i think
What would adding a lappack/batpack module even do? If you mean "make electric tools draw from the armor before drawing from themselves" you should say that.
If you have UE you can get infinite power in creative with the use of an infinite battery. I'm really suprised that idea worked, but it did, so if you need limitless creative power: there ya go. As in just, have an infinite battery in your inventory. Your power gauge turns into an infinity. Its totes boss.
Build #329 is now recommended!
-GUI now shows what modules are installed, as well as which ones are disabled in config.
-Module list can now be scrolled - small blue arrows will indicate when there is more on a page, and you can click the arrows or use the mousewheel to scroll. This mostly won't be noticed unless Andrew2448's addons is installed (which have had quite a few new additions!).
-Various bugfixes (item names, API mismatches)
-Custom models are go!
As promised, the visor and the crystals can be coloured separately from the main armor bits, and they glow in the dark! Extra bits and animations still to come, but this sets the groundwork for a variety of other .obj armor models if people want to make them.